Multiphase Modeling of a Flowing Electrolyte – Direct
Methanol Fuel Cell
by
David Ouellette
A thesis submitted to the Faculty of Graduate Studies and Research
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Mechanical Engineering
Ottawa-Carleton Institute for Mechanical and Aerospace Engineering
Department of Mechanical and Aerospace Engineering
Carleton University
Ottawa, Ontario, Canada
December, 2015
David Ouellette
c©Copyright
The undersigned hereby recommends to the
Faculty of Graduate Studies and Research
acceptance of the thesis
Multiphase Modeling of a Flowing Electrolyte – Direct Methanol
Fuel Cell
submitted by David Ouellette
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Dr. Edgar A. Matida, Co-supervisor
Dr. Cynthia A. Cruickshank, Co-supervisor
Dr. Metin I. Yaras, Chair,Department of Mechanical and Aerospace Engineering
Ottawa-Carleton Institute for Mechanical and Aerospace Engineering
Department of Mechanical and Aerospace Engineering
Carleton University
December, 2015
ii
Abstract
Direct methanol fuel cells (DMFCs) are considered one of the leading contenders for
low power applications due to their energy dense, liquid fuel as well as low greenhouse
gas emissions. However, DMFCs have lower than predicted performance due to methanol
crossover. One proposed solution is to allow a liquid electrolyte, such as diluted sulfuric acid,
to flow between the anode and cathode, thereby removing any methanol that attempts to
crossover to the cathode. The corresponding fuel cell is named the flowing electrolyte - direct
methanol fuel cell, or FE-DMFC. So far few researchers have examined the effectiveness of
this fuel cell and none have explored the multiphase flow within the membrane electrode
assembly (MEA) of this fuel cell.
In this study, the well-known Multiphase Mixture Model (MMM) was improved with a
new single domain approach which was used to model the flow behaviour and performance of
the FE-DMFC. Unlike the existing methods, the proposed model only requires the mixture
variables, thereby removing the requirement for information about the gaseous state, when
attempting to couple the porous and electrolyte layers together. Furthermore, the model’s
formulation gives the capability to resolve liquid saturation jumps in a single domain
manner. The proposed approach is sufficiently flexible that it could be applied to other
modeling methods, such as the Multi-Fluid Model (MFM). The corresponding derivation
for the MFM is provided. The fidelity of the improved MMM is examined through 3 test
cases, which include a comparison to: the analytical liquid saturation jump solution, the
analytical single phase solution for the FE-DMFC, and to in-house FE-DMFC experimental
iii
data. The numerical model was shown to be capable of accurately reproducing all three
test cases.
To understand the FE-DMFC’s underlying physics, the numerical model was applied
under baseline operating conditions, and a series of parametric studies were conducted to
understand the effect that: the anode and cathode membrane (AM and CM, respectively)
thicknesses and the flowing electrolyte channel’s (FEC) porosity and thickness each have
on the fuel cell’s performance. The findings of the parametric study were used to provide
recommendations on conditions which yield maximum power density and minimal methanol
and water crossover. The results from the baseline study suggest that the FE-DMFC is
capable of effectively reducing methanol crossover by at least 20 fold, when compared to the
DMFC. However it was found that the back pressure within the FEC is an important feature
to consider, as this can cause the bulk fluid to flow from the FEC to the anode and cathode
compartments, causing a counterflow condition. Although this aids in reducing methanol
crossover even further, it was found that the anode activation polarization also increased,
thereby reducing the fuel cell’s performance. The results from the parametric study suggest
that a thin AM and thick CM arrangement should be used; on the order of 88.9 μm and
177.8 μm respectively, corresponding to Nafion R© 1135 and 117 membranes respectively;
which is consistent with trends found in previous experimental studies. The results also
suggest that a fully open FEC (porosity of one) will provide the greatest performance.
Although this finding contradicts existing experimental data, considerations such as the
choice of catalyst layer wettability and back pressure within the FEC are provided to achieve
a membraneless FE-DMFC with a fully open FEC.
iv
To my mother and late father, Pierrette and Denis, and to my older but little sisters,
Céline and Sarah, for a lifetime of love, support and encouragement.
v
Acknowledgments
First and foremost, I would like to thank my supervisors, Dr. Edgar Matida and Dr.
Cynthia Ann Cruickshank, for giving me the opportunity to work on this project. Their
patience, support and guidance towards my work and their open door policy are greatly
appreciated. Whenever we met, they always displayed enormous enthusiasm and passion
towards teaching and research and I found it very contagious. I would also like to thank Dr.
Feridun Hamdullahpur for his support and generosity in the initial years of my work and
Dr. Glenn McRae for his invaluable input and for our many fruitful discussions in the area
of electrochemistry. We would frequently lose track of time during these discussions. Over
the years, he has been an incredible wealth of information in seemingly everything. I would
also like to thank the technologists and machinists at Carleton University for their help
and guidance in my experimental work. As well as Neil McFadyen for his help in giving me
access to the computational resources on campus and for his technical support.
I would also like to extend a special thanks to my friend and colleague Dr. Can Ozgur
Colpan from Dokuz Eylül University. His patience, invaluable guidance and expertise
in fuel cell modeling helped make this work possible. I would not be where I am today
without his help. I would also like to extend my thanks to the members of the Fuel Cell
Energy Systems Laboratory, both past and present, for their advice and discussions on fuel
cell modeling and experimentation. This includes: David Chan, Eric Duivesteyn, Yashar
Kablou, Prameela Karumanchi and Nasim Sabet-Sharghri. I wish you all the very best in
your undoubtedly bright futures. I would also like to thank my friends and past and present
vi
lab mates: Christopher Baldwin, Sébastien Brideau, Jayson Bursill, Jenny Chu, Ryan
Dickinson, Nina Dmytrenko, Phillip Droulliard, Ifaz Haider, Kenny Lee Slew, Mike Miller,
Patrice Pinel, John Polansky, and Adam Wills for all their help and advice concerning my
experimental and modeling work. They all helped me retain my sanity (or what’s left of it)
by periodically dragging me away from my work for a tea break.
I also thank the Ontario Centres of Excellence (OCE), Natural Sciences and Engineering
Research Council of Canada (NSERC) and the Ontario Graduate Scholarship (OGS) for their
financial support. Without their contributions, this work would not have been possible.
vii
Table of Contents
Abstract iii
Acknowledgments vi
Table of Contents viii
List of Tables xiii
List of Figures xiv
Nomenclature xx
1 Background 1
1.1 Flowing Electrolyte – Direct Methanol Fuel Cells . . . . . . . . . . . . . . . 1
1.2 Operating Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 FE–DMFC Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Backing Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Catalyst Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.3 Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.4 Flowing Electrolyte Channel . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Performance Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Polarization Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.2 Methanol Crossover . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
viii
1.4.3 Water Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Overview of Modeling Approaches . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 Two-Phase Modeling Approaches . . . . . . . . . . . . . . . . . . . . 16
1.5.2 Computational Domain . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.3 Liquid Saturation Distributions . . . . . . . . . . . . . . . . . . . . . 17
1.6 FE–DMFC Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6.1 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6.2 Modeling Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7 Unresolved Issues and Research Objectives . . . . . . . . . . . . . . . . . . . 23
1.8 Organization of Presented Thesis . . . . . . . . . . . . . . . . . . . . . . . . 25
2 Modeling Approach and Formulation 27
2.1 Computational Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Modeling Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3 Reduction of Governing Equations . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.2 Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.3 Conservation of Momentum . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.4 Conservation of Species . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.5 Conservation of Charge . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.4 Constitutive Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.1 Mixture Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.2 Electrochemical Relationships . . . . . . . . . . . . . . . . . . . . . . 42
2.4.3 Source Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Spherical Agglomerate Sub-Model . . . . . . . . . . . . . . . . . . . . . . . . 46
2.5.1 Catalyst Layer Porous Properties . . . . . . . . . . . . . . . . . . . . 49
2.6 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.7 Numerical Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
ix
2.7.1 Numerical Challenges and Stability Considerations . . . . . . . . . . 55
2.8 Grid Generation and Independence . . . . . . . . . . . . . . . . . . . . . . . 62
2.8.1 Grid Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.8.2 Grid Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3 Model Verification and Validation 65
3.1 Test Case 1 – Comparison to an Analytical Liquid Saturation Jump Model . 65
3.2 Test Case 2 – Comparison to an Analytical Single Phase FE-DMFC Model . 69
3.3 Test Case 3 – Comparison to Experimental FE-DMFC Data . . . . . . . . . 72
3.3.1 Fuel Cell Assembly and Design . . . . . . . . . . . . . . . . . . . . . 72
3.3.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3.3 Comparison between Experimental and Simulated Results . . . . . . 79
4 Modeling Case Studies 81
4.1 Baseline Performance Characteristics . . . . . . . . . . . . . . . . . . . . . . 81
4.1.1 Pressure and Velocity Distributions . . . . . . . . . . . . . . . . . . . 81
4.1.2 Liquid Saturation and Water Content Distributions . . . . . . . . . . 84
4.1.3 Liquid Methanol Concentration Distribution . . . . . . . . . . . . . . 87
4.1.4 Gaseous Oxygen Concentration Distribution . . . . . . . . . . . . . . 90
4.1.5 Cathode Catalyst Layer Performance . . . . . . . . . . . . . . . . . . 91
4.2 Effect of FEC Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Effect of FEC Porosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.4 Effect of Anode and Cathode Membrane Thickness . . . . . . . . . . . . . . 96
4.5 Comparison of DMFC and FE-DMFC Performance . . . . . . . . . . . . . . 100
5 Conclusions and Future Work 103
5.1 Conclusions and Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . . 105
x
List of References 107
Appendix A Derivation of the Liquid Saturation Equation for the Multi-Fluid
Model (MFM) 120
Appendix B Correlations and Properties Used for Modeling Studies 123
Appendix C Derivation of the Analytical Liquid Saturation Distribution for
Test Case 1 126
Appendix D Derivation of the Analytical Single Phase Model for Test Case
2 129
D.1 Modeling Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
D.2 Mass and Momentum Transport . . . . . . . . . . . . . . . . . . . . . . . . . 130
D.3 Methanol Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
D.4 Oxygen Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
D.5 Electrochemical Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . 133
D.6 Solution Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Appendix E Uncertainty Analysis for Test Case 3 135
E.1 Uncertainty Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
E.2 Elemental Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
E.2.1 Error Estimation in the Temperature Control System . . . . . . . . . 137
E.2.2 Error Estimation in the Fluid Control System . . . . . . . . . . . . . 137
E.2.3 Error Estimation in the Fluid Concentrations . . . . . . . . . . . . . 138
E.2.4 Error Estimation in the Active Area . . . . . . . . . . . . . . . . . . 139
E.2.5 Summarized Bias Uncertainty . . . . . . . . . . . . . . . . . . . . . . 139
E.3 Overall Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Appendix F Model Calibration Procedure 143
xi
F.1 Coordinate Search Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
F.2 Surrogate Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
F.3 Calibration Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
xii
List of Tables
2.1 Constitutive equations used to calculate the mixture fluid and transport prop-
erties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2 Summary of reaction-based source terms for each governing equation. . . . . 45
3.1 Summary of the boundary conditions, and geometric and porous properties
of the two layers used for Test Case 1. . . . . . . . . . . . . . . . . . . . . . 67
3.2 Mass loading of each component within the ACL and CCL. . . . . . . . . . . 74
3.3 Baseline operating conditions for the experiments with estimated bias errors. 77
3.4 Summary of the lowest, average and highest uncertainties for the experimental
measurements presented in Test Case 3. . . . . . . . . . . . . . . . . . . . . 78
B.1 Boundary conditions used for the baseline operating conditions in the pre-
sented modeling studies. The second set of subscripts under the the symbol
column represents the corresponding interfaces for that variable. . . . . . . . 123
B.2 Electrochemical and transport properties used in modeling study. . . . . . . 124
B.3 Fuel cell dimensions and material properties used in modeling study. . . . . . 125
F.1 Values of calibration parameters used in this work and their constraints en-
forced during the calibration process. . . . . . . . . . . . . . . . . . . . . . . 146
xiii
List of Figures
1.1 Schematic of a flowing electrolyte-direct methanol fuel cell (FE-DMFC). . . . 3
1.2 In-house SEM micrograph of carbon paper (Toray R© TGP-H-090) at 200×magnification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 In-house micrographs of the cathode catalyst structure. (a) Shows the meso-
scale structure observed through SEM, whereas (b) shows the nano-scale struc-
ture which forms the meso-scale structure, which was observed through TEM. 6
1.4 Chemical composition of a Nafion R© membrane . . . . . . . . . . . . . . . . . 7
1.5 In-house SEM micrograph of a 1.57 mm porous polyethylene spacer used for
the FEC at 100× magnification. . . . . . . . . . . . . . . . . . . . . . . . . . 81.6 Schematic of a typical fuel cell polarization curve, and the regions dominated
by each loss mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.7 Demonstration of the saturation jump across adjacent layers of differing
porous properties, for an assumed uniform capillary pressure of -10 kPa. The
black line, in the top right sub-figure, qualitatively shows the liquid saturation
distribution between the three layers. . . . . . . . . . . . . . . . . . . . . . . 18
2.1 Schematic of the computational domain used in the model. The layers en-
closed within the dashed box are the ones that are considered within the model. 28
2.2 Comparison of the capillary diffusion coefficient of water, DH2Ocap , and the
Young-Laplace diffusion coefficient of water, Dψcap, both at 80◦C and for a
porous media with ε = 0.78, K = 10−12 m2 and θc = 110◦. . . . . . . . . . . 37
xiv
2.3 Comparison of the diffusion coefficient proposed by Mortupally et al., and its
transformation into the effective diffusion coefficient of water, DH2Oλ,eff , both at
80◦C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4 Schematic of a spherical agglomerate. The agglomerate nucleus has a radius of
Ragg, while the electrolyte and water film thicknesses are δe and δl, respectively. 47
2.5 Boundary conditions used in the proposed model. The variables mentioned in
the figure are values that are set as a boundary condition for the corresponding
variable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.6 Schematic of a representative control volume and their dimensioning in the
x-direction. The same approach is used for the y-direction. . . . . . . . . . . 53
2.7 Flow chart showing the solution process for the proposed model. . . . . . . . 56
2.8 Convergence history for a typical simulation at 0.1 V. . . . . . . . . . . . . . 57
2.9 Sample pulse functions, with δ = 0.01, 1, 5, 10, 15 and 20 μm, and the
inflection points of the pulses are arbitrarily placed at 100 and 200 μm. . . . 61
2.10 Schematic of the grid refinement approach used in this work, for an arbitrary
material layer. Each grid layer has its own specified thickness and specified
number of body-fitted nodes. This refinement template is repeated for each
material layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.11 Comparison of baseline numerical results at different levels of grid refinement.
The results in a-c are for a cell voltage of 0.1 V, whereas the results in d and
within each sub-figure are for cell voltages from open circuit to 0.1 V. . . . . 63
3.1 Schematic of two adjacent layers with differing Young-Laplace coefficients,
where the mass flux at x = 0, and the liquid saturation at x = t1 + t2 are
known. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
xv
3.2 Comparison of the analytical (black line) and numerical (dashed coloured
lines) MMM solutions of the saturation jump equations, for the case where
Layer 1’s contact angle is varied as shown in the graphs. a) Displays the full
range of the two layers, whereas b) displays the solutions near the mating
layers’ interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3 Comparison of the analytical (black line) and numerical (dashed coloured
lines) MMM solutions of the saturation jump equations, for the case where
∇ψ = 0 in the numerical model and Layer 1’s contact angle is varied, asshown in the graphs. a) Displays the full range of the two layers, whereas b)
displays the solutions within Layer 1 and near the mating layers’ interface. . 68
3.4 Comparison between the analytical (black solid line) and numerical (coloured
dashed line) results, under bas line operating conditions and varied current
densities. The following figures display the following profiles: (a) polariza-
tion curves, (b) anode and cathode activation polarization, (c) gauge pressure
distributions, (d) mixture velocity distribution (note the change in units on
the secondary axis), (e) methanol concentration distribution, and (f) oxygen
concentration distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.5 Photograph displaying the layout of the FE-DMFC components (left) as well
as the fully assembled cell (right). . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Photograph of a half-MEA fastened to the graphite plate with a sheet of Teflon
tape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.7 Photograph of the experimental setup used in this study. . . . . . . . . . . . 76
3.8 Schematic of the experimental setup used for the operating fuel cell measure-
ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.9 Comparison between the experimental (data points) and numerical (solid
lines) results, for: (a) varied cell temperature, and (b) varied inlet methanol
concentration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
xvi
3.10 Normalized parity plot (i/imax) showing all experimental data points for the
(a) varied cell temperature experiments, and (b) varied inlet methanol con-
centration experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1 Modeled effect of current density on (a) mixture gauge pressure, and (b)
mixture velocity, both under the baseline conditions. . . . . . . . . . . . . . 82
4.2 Variation of mixture kinematic viscosity, ν, and mixture density, ρ, with re-
spect to liquid saturation at a temperature of 80◦C. . . . . . . . . . . . . . . 84
4.3 Modeled effect of current density on the liquid saturation and water content
distributions, under the baseline operating conditions. . . . . . . . . . . . . . 86
4.4 Modeled effect of current density on (a) the molar flux of water and its com-
ponents through the AM, and (b) molar flux of water and its components
through the CM, all under the baseline operating conditions. . . . . . . . . . 86
4.5 Modeled effect of current density on (a) the liquid methanol concentration,
and (b) the distribution of methanol crossover and its components, through
the CM under baseline operating conditions. . . . . . . . . . . . . . . . . . . 88
4.6 Modeled effect of current density on the methanol crossover current density
through (a) the anode membrane, and (b) cathode membrane, under base-
line operating conditions. Note that for legibility, both figures’ y-axes have
different scales. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.7 Modeled comparison of the fuel efficiency, when the removed methanol at the
FEC outlet is and is not recycled to the anode inlet. . . . . . . . . . . . . . . 90
4.8 Modeled effect of current density on the gaseous concentration of oxygen. . . 91
4.9 Modeled effect of current density on the CCL’s (a) agglomerate correction
factor and (b) agglomerate effectiveness factor. . . . . . . . . . . . . . . . . . 92
4.10 Modeled effect of current density on the water film thickness surrounding the
CCL agglomerates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
xvii
4.11 Modeled effect of the FEC’s thickness on (a) water crossover flux, and (b)
crossover current density. Crossover current densities calculated for tFEC > 3
mm were not shown, as they were deemed negligible. . . . . . . . . . . . . . 94
4.12 Modeled effect of the FEC’s porosity on (a) the average area specific resistance
and maximum power density of the fuel cell, and (b) the crossover current
density. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.13 The first two figures display the modeled effect of the AM thickness on the
average (a) anode and (b) cathode water crossover fluxes. Whereas the last
two figures display the modeled effect of the CM thickness on the average
water crossover flux in the (a) anode and (b) cathode water crossover fluxes. 97
4.14 Modeled effect of the (a) AM and (b) CM thickness on the average crossover
current density. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.15 Modeled effect of the individually varied AM and CM thicknesses on the
maximum power density. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.16 Modeled effect of the individually varied AM and CM thicknesses on the
maximum power density. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.17 Comparison of each fuel cell configuration’s modeled (a) crossover current
density and (b) area specific resistance. . . . . . . . . . . . . . . . . . . . . . 101
4.18 Comparison of the (a) FE-DMFC’s and (b) the modified DMFC’s liquid
methanol concentration profile. . . . . . . . . . . . . . . . . . . . . . . . . . 102
E.1 Averaged polarization curves during each experimental run and tested MEA,
with 4000 mol m−3 inlet methanol concentration. Each polarization curve
presented here is the average of the 5 polarization curves collected during
that experimental run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
E.2 Compiled experimental data with the uncertainties for polarization curves at
cell temperatures of 40◦C, 60◦C and 80◦C. . . . . . . . . . . . . . . . . . . . 141
xviii
E.3 Compiled experimental data with the uncertainties for polarization curves at
inlet methanol concentrations of 1000 mol m−3, 2000 mol m−3 and 4000 mol
m−3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
xix
Nomenclature
Abbreviations
ABL Anode Backing Layer
ACL Anode Catalyst Layer
AFC Anode Fuel Channel
AM Anode Membrane
CAC Cathode Air Channel
CBL Cathode Backing Layer
CCL Cathode Catalyst Layer
CM Cathode Membrane
DMFC Direct Methanol Fuel Cell
EOD Electro-osmotic Drag
FE-DMFC Flowing Electrolyte – Direct Methanol Fuel Cell
FE Flowing Electrolyte
FEC Flowing Electrolyte Channel
MEA Membrane Electrode Assembly
xx
MFM Multi-Fluid Model
MMM Multiphase Mixture Model
MOR Methanol Oxidation Reaction
NIST National Institute of Standards and Technology
OCV Open Circuit Voltage
ORR Oxygen Reduction Reaction
Pt Platinum
Ru Ruthenium
SEM Scanning Electron Micrography
TEM Transmission Electron Micrography
Variable
A Active Area m2
a Activity -
Bx Bias Error of Variable x -
C Concentration mol m−3
D Diffusion Coefficient m2 s−1
d Pore Diameter m
E Effectiveness Factor -
xxi
F Faraday’s Constant (96 485) C mol−1
f Mass Fraction/Interpolation Factor -/-
gf Gibb’s Free Energy of Formation J mol−1
h Mass Transfer Resistance s m−3
I Current A
i Current Density A m−2
ilim Limiting Current Density A m−2
J Leverett J-Function -
j Volumetric Current Density A m−3
jl Capillary Diffusion Flux for the Liquid Phase kg m−2 s−1
kc Agglomerate Reaction Rate s−1
kH Henry’s Constant -
kr Relative Permeability -
K Absolute Permeability m2
Ka Reaction Constant for Methanol Oxidation mol m−3
L Length m
M Molecular Weight kg mol−1
m′′ Mass Loading kg m−2
Ṅ Molar Flow Rate mol s−1
xxii
Ṅ ′′ Molar Flux mol m−2 s−1
n Number of Electrons Involved in Reaction -
nagg Agglomerate Density m−3
nd Coefficient of Electro-osmotic Drag -
P Pressure Pa
Pe Peclet Number -
R Radius of Whole Agglomerate / Electrical Resistance m / Ω m2
R̄ Universal Gas Constant (8.3144) J mol−1 K−1
Rx Random Error of Variable x -
Re Reynolds Number -
ragg Agglomerate Nucleus Radius m
Sgen Consumption/Generation Flux kg m−3 s−1
Skgen Consumption/Generation Flux of Species k mol m−3 s−1
Strans Transport Source Term kg m−3 s−1
s Liquid Saturation -
t Thickness m
tν,95% Student t-Distribution Factor at 95% Confidence -
T Temperature K
u Fluid Velocity in the x-Direction m s−1
xxiii
Vcell Cell Voltage V
v Fluid Velocity in the y-Direction m s−1
Q̇ Volumetric Flowing Electrolyte Flow Rate m3 s−1
x Position in the x-direction / Mole fraction m / -
y Position in the y-direction m
Greeks Letters
α Transfer Coefficient / Water Transport Coefficient -
γ Advection Correction Factor -
Γ General Diffusion Coefficient -
δ Thickness / Pulse Diffusion Index m / -
ε Porosity / Volume Fraction -
η Overpotential V
ηfuel Fuel Efficiency -
θc Contact Angle◦
ϑ Pulse Function -
Θ Objective Function -
κ Electrical Conductance S m−1
λ Mobility -
xxiv
λwc Water Content -
ν Kinematic Viscosity / Degrees of Freedom m2 s−1 / -
ξ Agglomerate Correction Factor -
ρ Density kg m−3
σ Surface Tension N m−2
τ Tortuosity -
Φ Electrical or Ionic Potential V
φ Arbitrary Variable -
φR Thiele’s Modulus at Position R -
ψ Young-Laplace Coefficient Pa
ω Relaxation Factor -
Subscript/Superscript
ABL Anode Backing Layer
ACL Anode Catalyst Layer
AFC Anode Fuel Channel
AM Anode Membrane
a Anode
act Activation Polarization
xxv
CAC Cathode Air Channel
CBL Cathode Backing Layer
CCL Cathode Catalyst Layer
CM Cathode Membrane
c Cathode
cap Capillary
conc Concentration Polarization
contact Contact Resistance
e Electrolyte Phase
eff Effective Value
eq Equilibrium Value
FEC Flowing Electrolyte Channel
g Gaseous State
H2O Water
in Inlet
k Species k
l Liquid State
lg Two-phase Condition
MeOH Methanol
xxvi
O2 Oxygen
o Intrinsic Property
ohmic Ohmic Polarization
ref Reference Value
rev Reversible Conditions
s Solid Phase
xover Crossover
- Averaged value
τ Tortuosity
xxvii
Chapter 1
Background
1.1 Flowing Electrolyte – Direct Methanol Fuel Cells
Low temperature fuel cells have shown great promise toward auxillary power, portable and
transportation applications; in particular, hydrogen and methanol fuel cells [1,2]. Hydrogen
is the ideal fuel for these applications due to its high energy density (119.9 MJ kg−1 at
300 bar), abundance and zero greenhouse gas emissions during operation [1]. This type
of fuel cell is commonly known as the proton exchange membrane fuel cell or PEMFC.
However, the challenges associated with the storage, transportation and distribution of
hydrogen have hampered the growth of this type of fuel cell. A promising alternative fuel
is methanol. Under standard conditions, methanol is in a liquid state, which mitigates
the challenges associated with fuel storage. Methanol yields an energy density that is
approximately 26× higher than hydrogen after consideration of storage efficiency. Thiscorresponds to an energy density of 18.9 MJ kg−1 for methanol and 0.72 MJ kg−1 for
hydrogen [1]. In addition, methanol is inexpensive, simple and potentially sustainable to
produce, and the infrastructure to support methanol is already present. This type of fuel
cell is commonly known as the direct methanol fuel cell or DMFC.
Unfortunately, DMFCs suffer from a phenomena known as ‘methanol crossover’, which
1
2
will be discussed in more detail in Section 1.4.2. It has been estimated that the losses
associated with methanol crossover account for 30% of all losses within this fuel cell [3]. As a
mitigation strategy, this fuel cell has been designed to operate with a decreased inlet methanol
concentration, typically less than 2 mol L−1 [4]; thus severely decreasing the DMFC’s energy
density. Hence the reduction of methanol crossover has been a major topic of research in
literature. One method to potentially solve this challenge is to use the flowing electrolyte
concept [5], where the corresponding fuel cell is known as a flowing electrolyte – direct
methanol fuel cell, or FE-DMFC. In the following section, the operating principles and the
individual components within the FE-DMFC will be discussed.
1.2 Operating Principles
The FE-DMFC is an electrochemical device which utilizes methanol as a fuel and oxygen or
air as an oxidant to generate electricity [1, 6]. A schematic of a typical FE-DMFC is shown
in Figure 1.1.
The generated electricity is achieved by supplying diluted methanol in the anode fuel
channel (AFC), which is then transported through a porous carbon backing layer (ABL)
and to a catalyst layer (ACL) – which is typically composed of platinum and ruthenium
(Pt and Ru) [7, 8]. The supplied diluted methanol then oxidizes in the ACL, which releases
electrons and protons, with carbon dioxide as the by-product. The completed anode reaction
is shown below the ACL, in Figure 1.1. The protons are transported through the anode and
cathode proton conducting yet electron insulating membranes (AM and CM) and flowing
electrolyte channel (FEC), which is typically composed of diluted sulfuric acid. The electrons
on the other hand, cannot conduct through the membranes and are therefore forced back
through the ACL, ABL and AFC to travel around an external circuit to be used towards
useful electrical work. The external circuit is connected to the anode and cathode current
3
HeCOOHOHCH 66223 OHHeO 22 36623
Flow
ing
Ele
ctro
lyte
Cha
nnel
(FE
C)
Cat
hode
Bac
king
Lay
er(C
BL
)
Cat
hode
Mem
bran
e(C
M)
Cat
hode
Cat
alys
tLay
er(C
CL
)
Ano
deB
acki
ngL
ayer
(AB
L)
Ano
deC
atal
ystL
ayer
(AC
L)
Ano
deM
embr
ane
(AM
)
H2SO4 + H2O + CH3OH
H2SO4 + H2O
O2
e- e-
Cat
hode
Air
Cha
nnel
(CA
C)
CO2
CH3OH + H2O
Ano
deFu
elC
hann
el(A
FC)
Ano
deC
urre
ntC
olle
ctor
(AC
C)
Cat
hode
Cur
rent
Col
lect
or(C
CC
)
H2O
e- e-
Figure 1.1: Schematic of a flowing electrolyte-direct methanol fuel cell (FE-DMFC).
collectors (ACC and CCC), as shown in Figure 1.1. On the cathode, oxygen is supplied to
the cathode air channel (CAC) and is transported through a porous carbon backing layer
(CBL) to the catalyst layer (CCL), which is typically composed of Pt. Here, the supplied
oxygen is reduced with the provided electrons and protons from the anode reaction, to form
water. The anode, cathode and overall reactions are summarized below.
Anode: CH3OH +H2O −−−−→ CO2 + 6H+ + 6e− Eeqa = −0.03 VCathode:
3
2O2 + 6H
+ + 6e− −−−−→ 3H2O Eeqc = 1.24 V
Overall: CH3OH +3
2O2 −−−−→ CO2 + 2H2O Eeq = 1.21 V
Since at a given operating point, not all methanol is consumed in the ACL, the remaining
methanol is transported through the AM, FEC and CM, and reacts within the CCL. This
causes methanol to rapidly oxidize in an oxygen rich environment, also causing the released
4
electrons and protons to be readily consumed by the surrounding oxygen. Since these
electrons do not travel through the external load, this results in lost work. This effect
is known as methanol crossover and will be discussed in more detail in Section 1.4.2.
However, the FEC in this fuel cell is designed to act as a methanol barrier, by removing any
crossed over methanol, thus protecting the CCL. The addition of the FEC could potentially
allow for less expensive membranes to be used, if the FEC is able to effectively remove
methanol from the fuel cell. The removed methanol could then be separated and supplied
to the anode inlet to be recycled [5], or sent to a specialized fuel cell that could use the
liquid electrolyte-methanol mixture as fuel [9]. As will be seen in Sections 1.6 and 1.7,
little is known about how the FEC affects the performance of the FE-DMFC. As such
the focus of this dissertation will be on understanding how this fuel cell functions, and
to lay a foundation for the fuel cell community, such that future work could focus on the
optimization of the fuel cell’s structure and design.
It should be noted that the DMFC, without the FE, only has one membrane and does not
have the FEC. This configuration is also true for the PEMFC, where in this case, hydrogen
is used as fuel instead of methanol. Since individual components within the FE-DMFC were
mentioned, the next section will be devoted to describing their structure and physical use.
1.3 FE–DMFC Components
As can be seen in Figure 1.1, the FE-DMFC is composed of an anode and cathode backing
layer (ABL and CBL), an anode and cathode catalyst layer (ACL and CCL), two mem-
branes one for the anode compartment (AM) and one for the cathode (CM), and a flowing
electrolyte channel (FEC). Each of these components and their functions will be discussed
in the following sections, beginning with the backing layers (BLs).
5
Figure 1.2: In-house SEM micrograph of carbon paper (Toray R© TGP-H-090) at 200×magnification.
1.3.1 Backing Layers
The backing layers are a porous layer, typically composed of carbon, that allows the uniform
distribution of fuel and oxidant to the catalyst layers. Often times, a hydrophobic mate-
rial, such as Teflon R©, is impregnated into the BLs to help manage the rate of transport of
reactants. This in turn helps to reduce the rate of methanol crossover, and the removal of
liquid water from the cathode [4]. Under a scanning electron microscope (SEM), shown in
Figure 1.2, the carbon paper is shown to be a fibrous porous media. The webbing between
fibers is the impregnated Teflon R©.
1.3.2 Catalyst Layers
The catalyst layers are where the methanol oxidation reaction (MOR) in the anode and the
oxygen reduction reaction (ORR) in the cathode occur. In FE-DMFCs, the catalysts are
typically composed of a platinum-ruthenium (Pt-Ru) binary catalyst for the anode and Pt
catalyst for the cathode. The binary catalyst allows for improved catalytic activity, while the
Ru helps to protect the Pt from impurities and intermediate reactants from being absorbed
onto its surface [10, 11], which is known as catalyst poisoning. Like the BLs, the CLs are
sometimes impregnated with Teflon R© to increase the layer’s hydrophobicity [4]. Figure 1.3a
6
Figure 1.3: In-house micrographs of the cathode catalyst structure. (a) Shows the meso-scale structure observed through SEM, whereas (b) shows the nano-scale structurewhich forms the meso-scale structure, which was observed through TEM.
displays a SEM micrograph of the surface of the CL applied to a Toray R© TGP-H-090 backing
layer. As can be seen in the SEM micrograph, shown in Figure 1.2a, it can be seen that
the catalyst agglomerates into spherical pellets. Each pellet is further composed of smaller
agglomerates of composed of carbon, Nafion and catalyst, as shown in Figure 1.3b.
1.3.3 Membranes
The membranes used (most commonly Nafion R©) allow the passage of protons from the ACL
to the CCL, while forcing electrons to flow around an external circuit to power an external
load [12]. As such, this layer must be proton conducting but not electron conducting.
Furthermore, the membranes act as a separator for the fuel, oxidant and liquid electrolyte;
preventing methanol, water and oxygen from crossing over and reacting on the opposite
side of the fuel cell. This membrane uses polytetrafluoroethylene (PTFE) as the structural
support and several side chains typically terminated with sulfonic acid groups (SO−3 ) to
promote proton conductivity [13]. Figure 1.4 displays a schematic of the Nafion R© structure.
Since these sulfonic acid groups (SO−3 ) are negatively charged, the protons generated
7
C
C
CCC
F
FF
FFFF
F FO
C CF3F
O
C FF
C FF
SO
O
O- H+
n
m
m
Sulfonic Acid(SO3-H+)
Polymer SideChain
HydrophobicPTFE Backbone
m = 1n = 6 ~ 10
Figure 1.4: Chemical composition of a Nafion R© membrane. Adapted from Jiao and Li [13].
from the MOR on the anode become attracted to the sulfonic acid groups and are thus
transported to the cathode. On the other hand, due to Nafion R©’s very low electronic
conductivity compared to the carbon in the BLs and CLs, the electrons generated within
the anode are forced around the external circuit, to the cathode. This creates a proton
conducting yet electron insulating membrane [13].
The transport of protons through Nafion R© occur in multiple modes. The first is related to
the proton’s weak attraction to the membrane’s sulfonic acid groups, which allow protons to
easily bond, detach and bond again with the next nearest sulfonic acid group. This proton
‘hopping’ motion, from one sulfonic acid group to the next, is known as the Grotthuss
mechanism [13, 14]. The other mode of transport is caused by protons bonding with water
molecules to create a hydronium-type ion, in the form of (H2O)mH+, and is carried with the
flow of water within the membrane. This transport mechanism is known as the vehicular
transport mechanism. The third mode of transport is through electro-osmosis, which occurs
due to the presence of an electric field between the anode and cathode. This causes the
protons to migrate from the anode to the cathode.
8
Figure 1.5: In-house SEM micrograph of a 1.57 mm porous polyethylene spacer used forthe FEC at 100× magnification.
1.3.4 Flowing Electrolyte Channel
The flowing electrolyte channel (FEC) is the space between the AM and CM for the liquid
electrolyte. The original concept for this channel was simply an open space between the
membranes [5]. However, in Sabet-Sharghri et al.’s work [15], they demonstrated that an
open channel design for the FEC was not practical for use, since the membranes can bulge
and come in contact during operation; thus hindering the flow of the electrolyte and decrease
the performance of the fuel cell. Instead, they proposed the use of a porous spacer, composed
of polyethylene. A SEM image of this layer is shown in Figure 1.5.
1.4 Performance Characterization
DMFC performance is typically characterized by measuring the maximum current and power
densities, as well as all loss mechanisms, which include: the activation, concentration and
ohmic polarizations, as well as methanol and water crossover fluxes. The simplest and most
frequently adopted approach, to qualitatively compare fuel cell performance, is through use
of polarization curves. This method will be discussed next sub-section, while the remaining
loss mechanisms will be discussed after.
9
Current Density [A m-2]
Cel
lVol
tage
[V]
Reversible Cell Voltage
Ohmic Polarization Dominated
Concentration Polarization Dominated
Activation Polarization Dominated
Mixed Potential Dominated
Figure 1.6: Schematic of a typical fuel cell polarization curve, and the regions dominatedby each loss mechanism.
1.4.1 Polarization Curves
A common method to qualitatively determine the fuel cell performance, is through use of a
voltage-current plot, known as a polarization curve. The advantage of polarization curves
is that they provide a quick indicator of the qualitative magnitude of the different loss
mechanisms within the fuel cell, such as the: mixed potential, and the activation, ohmic
and concentration polarizations. These dominant regions are shown in Figure 1.6.
Many models reproduce a polarization curve by subtracting the reversible cell voltage,
Vrev, by all the loss mechanisms, η, as shown below. Details of each of the remaining terms
will be discussed in the following subsections.
Vcell = Vrev − ηact − ηohmic − ηconc (1.1)
10
Reversible Cell Voltage
The reversible cell voltage, Vrev, is the maximum voltage that a single cell can achieve to
produce useful work, as shown in Equation 1.2. This equation is calculated based on the
change in Gibb’s free energy of formation, Δḡf , of the overall reaction (as mentioned in
Section 1.2). Here, n and F are the number of electrons participating in the overall reaction,
and Faraday’s constant, respectively. At 25◦C, Vrev = 1.21 V [1,16,17].
Vrev = −ΔḡfnF
(1.2)
As will be discussed later in greater detail, there is typically significant methanol crossover
at open circuit, which creates a mixed potential. This significantly reduces the fuel cell’s
Vrev, typically to a range of 0.5 V - 0.7 V [18,19], as seen in Figure 1.6. This drop in Vrev is
a combination of a decrease in magnitude of Δḡf , due to the shift in the cathode’s half-cell
reaction, and due to the increased cathode activation polarization, ηc [18, 20, 21].
Activation Polarization
The activation polarization corresponds to the potential difference, across the electrolyte-
catalyst particle interface, required to drive the reaction. At low current densities, this loss
mechanism is generally the most dominant, as can be seen in Figure 1.6. A common method
to predict this loss mechanism is the Butler-Volmer expression, shown in Equation 1.3 [22,23].
j = aio,ref
[〈∏k
(Ck
Ckref
)mk〉exp
(αaF
R̄Tηact
)−〈∏
k
(Ck
Ckref
)nk〉exp
(−αcF
R̄Tηact
)](1.3)
In this equation, j represents the measured volumetric current density, a is the surface
area to volume ratio of the electrode, io,ref is the exchange current density of the electrode
at the reference state, C is the molar concentration of the fuel or oxidant, m and n are the
reaction order of the anodic and cathodic reactions respectively, α is the charge transfer
11
coefficient, T is the temperature, and ηact is the activation overpotential. The subscripts, a
and c, represent the anode and cathode, respectively, whereas k represents the kth species
participating in the reaction.
As more current is generated by the fuel cell, the reactions are pushed more in the
forward direction, causing the terms representing the reverse reaction to become negligible.
This form of the equation is known as the Tafel equation, shown below for the anodic and
cathodic reactions, respectively.
j = ajo,ref
〈∏k
(Ck
Ckref
)mk〉exp
(αaF
R̄Tηact
)(1.4a)
j = −ajo,ref〈∏
k
(Ck
Ckref
)nk〉exp
(−αcF
R̄Tηact
)(1.4b)
It should be noted that the form of the Butler-Volmer and Tafel equations shown so far
assume that the concentration of ions at the electrode surface is at equilibrium with the bulk
solution, and that the reaction mechanism is a one-step and one-electron process [22, 23].
Although both the methanol oxidation and oxygen reduction reactions (MOR and ORR) do
not follow this, the Tafel expression is often applied due to its simplicity. A more accurate
and frequently adopted approach to account for the multi-step reaction of the MOR is the
one proposed by Meyers and Newman [24], shown below. There are many similar expressions
in literature, but all have the same general form [25, 26]. Here, the second set of terms in
the denominator account for the smooth transition from zero order to first order kinetics,
where before, in Equations 1.3 and 1.4a, a step change in reaction order is often taken when
changing from zero to first order kinetics [21, 27].
j =
ajoaCMeOHl exp
(αaF
R̄Tηact
)
CMeOHl +Kaexp
(αaF
R̄Tηact
) (1.5)
12
More rigorous approaches also exist for the MOR and ORR, which account for more
detailed reaction pathways and the change in available reaction area due to the absorbed
species onto the catalyst surface [7, 18, 28, 29]. However, Equations 1.4b and 1.5 seem to
be the most frequently adopted equations to represent the ORR and MOR, respectively.
Another common approach is to develop a set of modified equations that account for the
physical structure of the CLs. For example, the CLs could be treated as a packed bed of
spheres, as can be seen in Figure 1.3, and from there an analytical solution [30–32], or a
more detailed numerical solution [33,34], could be devised to account for the mass transport
resistance of reactants. This modification has been demonstrated to be an accurate physical
representation of experimental results.
Ohmic Polarization
The next dominant regime is from the ohmic polarization. This loss is due to the resistance
to electron and proton conduction through each layer of the fuel cell, as well as the contact
resistances between layers [35]. Models often use Ohm’s Law to predict this loss, as shown
below. This equation often displays a nearly linear trend with current density, any deviation
from this trend would stem from non-linearities in the resistance of each layer.
ηohmic = iR (1.6)
Concentration Polarization
The final loss mechanism occurs when the fuel cell becomes transport limited; whereby,
reactants cannot reach the CLs fast enough to maintain the reaction rate. As such, the
concentration polarization, ηconc, only becomes important near the limiting current density,
ilim, (the maximum achievable current by the fuel cell) and causes the sharp drop in perfor-
mance as observed in Figure 1.6. However, DMFCs generally do not reach their ilim, due
to their high activation and ohmic polarizations. As such, the concentration polarization
13
can be neglected with minimal loss in accuracy. If one wanted to model the effects of the
concentration polarization, the Nernst equation can be applied where the final form of the
equation is shown below [22].
ηconc =R̄T
nFln
(1− i
ilim
)(1.7)
1.4.2 Methanol Crossover
In DMFCs, the anode and cathode are separated by a membrane. Although this layer is
designed to act as a separator to methanol and oxygen, some unreacted methanol is trans-
ported through the membrane and reacts in the CCL; this is known as methanol crossover.
This process effectively short circuits the fuel cell, since the released proton and electrons
from the oxidized methanol in the CCL, readily bond with the available oxygen. As such,
the released electrons do not provide useful electrical work for the external load. There are
other issues surrounding methanol crossover, some of the main issues are mentioned below.
• Methanol crossover wastes fuel and also causes higher rates of oxygen consumptionand water production in the CCL [18, 21]. The increased oxygen consumption could
lead to oxygen starvation in the CCL, decreasing the overall performance of the fuel
cell. The higher water generation rate causes water to fill the CBL and CCL pores,
hindering oxygen’s ability to reach the CCL and its ability to react on the catalyst
surface, thus decreasing the performance of the fuel cell. Furthermore, the generated
water can travel into the CAC, causing channel blockage, where higher back pressures
are required to push water out of the channels. The issues surrounding the increased
water production and crossover will be discussed in the next sub-section.
• Oxidized methanol will form an oxide film on the Pt catalyst particles within the CCL,increasing the activation polarization and thus decreasing the performance of the fuel
cell [18, 24, 25].
14
• To compensate for the lower performance, higher catalyst loadings (typically Pt) areoften required in the CCL, greatly adding to the cost of the fuel cell [36, 37].
There are many mitigation strategies to minimize or eliminate the effects of methanol
crossover, such as: the use of selective catalysts and membranes [38, 39], as well as the use
of transport barriers such as micro-porous layers (MPLs) [40, 41]. Each of these approaches
have their merits. For example, selective cathode catalysts could allow the cathode to be, in
some sense, immune to methanol crossover. However, research on these catalysts are still in
their infancy these catalysts typically display lower catalytic activity towards oxygen than
the traditional Pt-based catalysts [42,43]. Also, since methanol is not consumed in the CCL,
any crossed over methanol would fill the CCL’s pores preventing oxygen from reaching avail-
able reaction sites. Selective membranes could selectively prevent methanol from entering
the membrane, essentially eliminating methanol crossover all together. However, these mem-
branes are often more expensive than commercially available membranes, and also generally
display higher ohmic resistance, lower chemical and thermal stability, and durability [38,39].
Transport barriers, such as MPLs, would allow the use of traditional materials, as well as
help control the motion of the gas and liquid phases, and the species within the fuel cell. The
use of MPLs and their design have also been demonstrated to improve fuel cell performance.
But this approach only minimizes methanol crossover, not eliminate it.
1.4.3 Water Management
The water management in portable DMFCs and stacks is very important. On the one
hand, water helps to promote high membrane conductivity [13]. However, too much water
can lead to cathode flooding; where liquid water occupies the cathode’s pores preventing
oxygen from reaching the CCL. Further complications arise when the liquid water reaches
the CAC, and large back pressures are required to remove the condensed water. This causes
maldistribution of reactants in the cathode, erratic fuel cell performance and in some cases
local membrane dehydration. Each of which complicate the ability to collect repeatable
15
experimental results [44]. Cycling of the membrane hydration can also lead to pin hole
formation, which could lead to device failure and, as discussed in the previous subsection,
very high rates of methanol crossover and flooding [45].
Typical water management approaches involve the use of custom material properties for
the BLs and CLs [45], the use of MPLs [45, 46], and specialized membranes, control of the
cathode inlet relative humidity [44], and use of various channel geometries [47]. To the
author’s knowledge, a water balance study has not been performed on the FE-DMFC.
1.5 Overview of Modeling Approaches
Early DMFC modeling work assumed the fuel cell operated under single phase conditions.
This assumption implies that all species in the anode remain in the liquid state while all
species in the cathode remain in the gaseous state. This is only realistic at low current den-
sities and crossover rates [48,49]. Although simplified, these assumptions allowed for simple
yet relatively accurate models to be formulated. Often, depending on the complexity of the
model, an exact solution could be devised [26, 50]. Although single phase models are good
for determining general trends, they also tend to over-predict the performance of the fuel cell.
Over time, as computational power increased and fuel cell models became more advanced,
multiphase models began to be formulated [21, 41, 51–60]. However, when methanol is oxi-
dized, carbon dioxide is produced. This creates a two-phase flow within the anode, since car-
bon dioxide is in the gaseous state, and methanol and water both co-exist in the gaseous and
liquid states. Furthermore, in the cathode, as oxygen is reduced and crossed-over methanol
is oxidized, water is generated; which creates a two-phase flow in the cathode. These ef-
fects are not taken into account in the single phase models, and are therefore a major short
coming. Since these effects are not captured, the predicted performance is overestimated.
16
1.5.1 Two-Phase Modeling Approaches
Three common multiphase modeling approaches exist in the fuel cell literature: unsatu-
rated/saturated flow theory (UFT and SFT) [41, 52, 53], multi-fluid model (MFM) [54–56]
and multiphase mixture model (MMM) [21, 57–60]. The UFT and SFT models assume
that the pressure gradient of the primary phase (gas phase for UFT, and liquid for SFT)
is negligible [52]. This allows for a simplified set of momentum equations to be solved,
allowing for a more direct calculation of the capillary pressure distribution, which drives
the flow in a porous media. This assumption begins to break down as the secondary phase
generation is high, which typically occurs at high current densities.
A more generalized formulation is the MFM, which accounts for the variation of both
phase pressures [54–56]. This modeling approach is frequently used in both DMFC and
PEMFC models due to its flexibility and accuracy. This approach uses a set of governing
equations for each phase, and each of the sets of governing equations interact through source
terms. Although this approach is more physically representative of the fluid flow, it is
also very non-linear due to the large number of differential equations that are required to
be solved, and due to their strong coupling. A recent approach to reduce the number of
differential equations in the MFM, is the mathematically equivalent MMM. This approach
treats the multiphase flow as an effective single phase mixture [57,61]; which is achieved by
taking each governing equation and summing it across all phases. A set of algebraic equations
are then used to ensure proper coupling between each phase and the mixture. For further
discussion on the comparison between the presented approaches, the reader is referred to
Wang and Chen [62]. The reader is also referred to Mazumder and Cole’s derivation [63,64]
and Sui et al.’s [65] extension to their work for a good review of the governing equations
required to evaluate the performance of a fuel cell. In this dissertation, the MMM is applied
and, wherever applicable, extensions and parallels to the MFM will be provided.
17
1.5.2 Computational Domain
When modeling the fuel cell domain, there are two common approaches, the multi-domain
approach [54, 66, 67] and the single-domain approach [68–70]. The multi-domain approach
models each layer within the fuel cell separately and the adjacent layers are coupled through
interfacial conditions (typically through a continuity and Dirchlet conditon) [54]. It is also
common to tailor each set of governing equations to each layer. It is expected that this
approach is more computationally stable but likely requires a greater number of iterations
to achieve convergence. The single-domain approach, on the other hand, treats the entire
fuel cell as a unified domain, removing the need for interfacial conditions [71]. In this case,
the governing equations are formulated in a manner that makes them independent of which
layer of the model is being examined. This approach is very convenient for fuel cell modeling.
To achieve a single-domain model, one needs to solve each governing equation with a con-
sistent variable. However, this can become problematic for the water’s species conservation
equation [58,72], where in the porous layers (BLs and CLs) the liquid saturation (the ratio of
liquid volume within the pore volume, given by s = V– l / V– pore) is often used as the variable
of choice. However, in the electrolyte phase (such as in the CLs and membrane), the liquid
saturation is not very meaningful, due to the hydrophilic membrane pores. This causes the
membranes, to a good approximation, to be completely liquid saturated [73]. As such, it is
more meaningful to rather track the water content (λwc, the number of water molecules per
sulfonic acid group within the membrane) or the concentration of water in the electrolyte
phase, CH2Oe [31,55,65]. This change in variables quickly demonstrates how a single domain
approach can be difficult to formulate.
1.5.3 Liquid Saturation Distributions
In a fuel cell assembly, known as a membrane electrode assembly (MEA), it is typical
to have different porous properties for each layer. In a two-phase flow regime, if two
18
-20
-15
-10
-5
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Cap
illar
y Pr
essu
re [k
Pa]
Liquid Saturation
MPL: K = 10-13 m2 ɛ = 0.30 θc = 120°
BL: K = 10-12 m2 ɛ = 0.80 θc = 110°
CL: K = 10-14 m2 ɛ = 0.30 θc = 95°
BL MPL CL
Figure 1.7: Demonstration of the saturation jump across adjacent layers of differing porousproperties, for an assumed uniform capillary pressure of -10 kPa. The black line, inthe top right sub-figure, qualitatively shows the liquid saturation distribution betweenthe three layers.
materials of differing porous properties are adjacent, one layer will act more like a ‘sponge’,
causing more liquid to be retained in one layer than the other [74]. This is also shown
graphically in Figure 1.7, for a uniform capillary pressure, Pcap, of -10 kPa across three
layers. Because each layer has its own Pcap – s curve, a jump in s will occur across
each interface; where for example, s is 0.96 in the BL, 0.17 in the MPL, and 0.50 in the
CL. Similar behaviour can be expected in the case where there is a spatial distribution
of porous properties, such as when any of the layers are compressed during fuel cell assembly.
To model this phenomena, two common methods in the literature are discussed. In the
first, which is more applicable to the MFM approach, the liquid and gaseous pressures can be
obtained from the conservation of mass and momentum equations. The difference in phase
19
pressures yield the Pcap distribution. The corresponding s distribution can be obtained from a
Pcap – s expression, such as the Young-Laplace equation, as shown in Equation 1.8 [55,56,75].
Here, σ is the surface tension, K is the absolute permeability, ε is the porosity, θc is the
contact angle, and J is the Leverett J-Function, which is a function of s.
Pcap = Pg − Pl = σ(cos θc)( εK
)1/2J (1.8)
The other approach is to obtain a governing equation (either from the liquid continuity
equation or an appropriate species equation) that is in terms of s [21,52,53]. A common form
of the liquid continuity equation that is used in such an approach is shown below, which is
derived from coupling Darcy’s Law for both the liquid and gas phases [76–78].
∇ ·[Kkrlνl
∂Pcap∂s
(∇s) + krlνgkrgνl
(ρgug)
]= Sgen,l (1.9)
This approach yields identical results as the previous approach when the spatial variation
in porous properties are negligible. Due to this assumption, Equation 1.9 is not capable
of capturing a liquid saturation jump. To circumvent this issue, many authors apply an
interfacial continuity and capillary pressure equality to the relevant interfaces [67, 79]. Due
to this interfacial condition, this approach is not well-suited for a single domain approach.
1.6 FE–DMFC Literature Review
This section is devoted to summarizing the experimental and modeling studies currently
published in the literature on the FE-DMFC. This section will then tie into the next section,
where the unresolved issues in FE-DMFC literature will be summarized, along with this
dissertation’s research objectives.
20
1.6.1 Experimental Studies
Kordesch et al. [5] demonstrated the FE and circulating electrolyte (CE) concepts for a
PEMFC and DMFC, by comparing the open circuit voltages (OCV) of these fuel cells
when the FEC was active and inactive. In their study, they found that the OCV improved
when the FEC was active and provided possible improvements on the FE-DMFC and plant
designs. A notable recommendation was the inclusion of a liquid electrolyte-methanol
separator in the plant design, such that the removed methanol and liquid electrolyte can
both be recycled. Later, Sabet-Sharghi et al. [15, 80] performed a set of parametric studies
on the FE-DMFC, where they investigated the effect of the: FEC’s inlet concentration,
flow rate and thickness, cell temperature, and methanol concentration and flow rate on the
FE-DMFC’s performance. They demonstrated that the FE concept indeed decreased the
amount of methanol reaching the CCL; however careful consideration had to be made to
the FEC design. A major finding was that an open FEC was impractical for use, since
the membranes can bulge and come into contact, hindering the performance of the FEC.
Instead, Sabet-Sharghi et al. suggested the use of a porous spacer to allow the FE to pass
through, while preventing the membranes from coming into contact.
Kablou also recently examined the performance of a five cell, FE-DMFC short stack
where he applied the advice of Sabet-Sharghi et al. [81]. In his experimental work Kablou
examined the effects of: stack temperature, methanol concentration and flow rate, and FE
and air flow rate. Kablou found that the FE concept increased the performance of the stack
and also found that at high operating currents, the cell voltage will drop significantly if
the FE were to stop flowing, which could be an effective method to rapidly shut down the
stack. He further determined that the FE is an effective method to maintain a uniform
temperature across the stack.
Duivesteyn experimentally measured the porous properties (particle diameter, porosity
21
and permeability) of different FEC materials, and their affect on the fuel cell’s performance,
when the cell temperature and the FE flow rate were varied [82]. Although Duivesteyn
reported inconsistent performance on the optimal FE flow rate, his results seem to indicate
that higher pore diameters and thus higher permeabilities require higher FE flow rates to
maintain optimal performance. Duivesteyn suspected that the FEC’s back pressure played
a significant role in the fuel cell’s performance.
Ouellette et al. [9] recently proposed to replace sulfuric acid with a MOR intermediate,
specifically formic acid. This design allowed for the FEC outlet to be routed to the anode
inlet, thereby readily recycling the formic acid-methanol mixture as fuel, removing the need
for a separator as suggested by Kordesch et al. [5]. Interestingly, they found that the formic
acid electrolyte DMFC (FAE-DMFC) had a maximum power output at a temperature of
48◦C, which could make this fuel cell a viable solution for hand held and portable appli-
cations. Although the fuel cell’s performance was limited due to the low conductivity of
formic acid, Ouellette et al. encouraged the exploration of different MOR intermediates that
could be used as a sulfuric acid substitute, as well as the use of conductivity enhancement
techniques such as dissolving a salt of strong acid into the electrolyte.
1.6.2 Modeling Studies
One of the first models developed for the FE-DMFC was by Kjeang et al. [83, 84]. In their
work, they developed a two and three dimensional (2 and 3D) computational fluid dynamic
(CFD) model to examine the theoretical effectiveness of an open, single-phase FEC. They
found, under their examined conditions, that the FEC can reduce the methanol crossover
flux by as much as 90%, and that the methanol crossover flux was inversely proportional to
the flow rate of the FE.
Colpan et al. recently developed an extension to Kjeang et al.’s analysis by creating a
22
1 and 2D single phase performance model for the DMFC and FE-DMFC [35, 85]. In these
studies, they conducted several simulations on the FE, methanol and air flow rates, FEC
thickness, and inlet methanol concentration. They found that the greatest gross power
output can be achieved with the highest possible FE flow rate and thinnest possible FEC,
and that higher power densities can be achieved with increased methanol and air flow
rates. However, the electrical efficiency was found to decrease with increased methanol flow
rate, since less methanol was available to be consumed. They also found that with the CE
concept, the efficiency of the fuel cell can be increased by as much as 57%. It should be
noted that Colpan’s DMFC models were compared against experimental DMFC data, and
that the FE-DMFC model was treated as an extrapolation of the results. This was due to
the lack of FE-DMFC experimental data at the time. Furthermore, the effects caused by
the FEC’s back pressure, to maintain its flow rate was not included.
Kablou also performed a two-phase hydrodynamic study of the anode flow field that was
coupled to an analytical single phase model within the MEA [81]. Kablou’s hydrodynamic
algorithm applied the Hardy-Cross algorithm, with a separated flow pressure drop and
void fraction method. Using his algorithm, Kablou tested the effect of cell temperature,
current density, and methanol concentration and flow rate on the hydrodynamics of the
single cell and stack anode flow fields. Kablou demonstrated the importance of the inclusion
of two-phase flow within the anode flow field, as well as displayed the differences in the
predicted solution when different pressure drop and void fraction methods were applied.
Out of the tested methods, the Lockhart-Martinelli and CISE pressure drop and void
fraction methods most accurately reproduced the experimental data. He also pointed out
that the two-phase gravitational pressure was the most dominant component of pressure,
and the greatest pressure drop occurred near the anode inlet manifold.
Later, Duivesteyn et al. [86,87] extended Kjeang’s work by examining the isothermal and
23
non-isothermal hydrodynamics and effectiveness of a porous single phase FEC. Their studies
approximated the FEC’s working fluid as water, and they did not account for the mass
transfer in the anode or cathode, and they focused on the pressure drops and flow behaviour
in such a channel for different porous structures. Duivesteyn et al. concluded that Darcy’s
Law, and as such a 1D velocity profile, was a valid approximation of the flow behaviour
within the FEC and that preheating the FEC has negligible impact on the flow behaviour
due to the very small thermal entry length. Duivesteyn et al. also provided recommendations
on the porous structure to minimize methanol crossover and pressure drop, such as: the use
of a porosity in the range of 0.5 – 0.6 and a particle diameter that is an order of magnitude
smaller, or greater, than the FEC thickness.
1.7 Unresolved Issues and Research Objectives
Reviewing the information discussed in Sections 1.5 and 1.6, several unresolved issues have
been brought up, and are summarized below.
1. Current FE-DMFC models are only single phase [35, 83–87]; with the exception of
Kablou’s two-phase hydrodynamic model [81]. It has been demonstrated in DMFC
and PEMFC literature, that the multiphase flow significantly affect the performance
of the fuel cell in a detrimental manner [88, 89]. It is not known how the multiphase
flow will affect the performance of the FE-DMFC.
2. The removal of methanol from the FEC has been accounted for by a convective sink
term, without the consideration of how its pressure and velocity would affect the rest of
the fuel cell [35,85]. The increased pressure within the FEC was further demonstrated
in Duivesteyn et al.’s work [86, 87]. It is suspected that this increased pressure could
cause a counter convective flux from the FEC to the anode and cathode. However, this
phenomenon and its affect on the fuel cell’s performance have yet to be explored.
24
3. The validation process for FE-DMFC performance models have consisted of comparing
against an analytical solution [83,84], or by producing an equivalent DMFC model and
comparing the results to DMFC experimental data [35, 85]. As mentioned previously,
it is suspected that the FE may affect the fuel cell’s performance in a non-linear fashion
and still has yet to be fully understood. As such, a more thorough verification and
validation process will be required to ensure accurate predictions.
4. Published experimental studies on the FE-DMFC have only qualitatively discussed
how the ohmic resistance, and the methanol and water crossover fluxes affect the
performance of the fuel cell [15, 81, 82]. However, in some cases, little theoretical
backing to the experimental observations have been provided. As such, further details
of the physics during fuel cell operation still need to be resolved.
5. There are few single domain models in literature that apply the MMM. The ones that
exist often use information concerning the gaseous state to determine the activity, and
thus the water content, within the membranes, complicating the model. A formulation
that makes use of mixture variables only would be more practical, and to the author’s
knowledge, has not been formulated.
6. A common method to determine the saturation profile is through application of Equa-
tion 1.9. However, in this form, an explicit set of interfacial conditions would be
required, as discussed in Section 1.5.3, in cases where saturation jump is to occur.
This explicit treatment is not useful for a single domain approach, as the saturation
jump should occur on its own, without the need of any explicit treatment. A single
domain approach to account for this effect requires formulation.
Of these unresolved issues, the most salient points to be raised are as follows:
• The current state of understanding of how the FE-DMFC functions is still in its infancy,and little is known as to how the FEC affects the performance of the fuel cell;
25
• The current modeling approaches need to be improved to increase the fidelity of thepredictions and ease of implementation; and,
• Better verification and validation practices need to be implemented, such that thesetools can provide better predictions and understanding of how the FE-DMFC functions.
As such, the objectives of this dissertation is to address these unresolved issues by devel-
oping a single domain MMM of the FE-DMFC. The derivation of this model is intended to
extend the MMM formulation, such that only mixture variables are required throughout the
computational domain, and to provide an approach to account for liquid saturation jumps
in a single domain manner. A focus of the study will be to demonstrate the capabilities and
limitations of this model, through verification and validation, as well as apply the model to
understand in detail how the fuel cell functions and how the FEC affects the performance of
the fuel cell.
1.8 Organization of Presented Thesis
This dissertation is separated into five chapters to achieve the objectives mentioned in the
previous section. A summary of each chapter is provided below.
Chapter 1 presents the necessary background information and literature review to un-
derstand the information in this dissertation. The scope and outline of this
dissertation are provided.
Chapter 2 details the derivation and formulation of a 1D, multiphase and single domain
model of the FE-DMFC. A detailed discussion on the assumptions and
their expected impact on the model’s predictions, along with the solution
methodology, numerical challenges, and calibration procedure are provided.
26
Chapter 3 presents three test cases that are used to verify and validate the model
presented in Chapter 2. Two test cases involve comparisons to analytical
solutions, while the other test case involves a comparison to in-house experi-
mental FE-DMFC data. Details of the analytical solutions and the collected
experimental data are provided in Appendices C, D and E, respectively.
Chapter 4 demonstrates the capabilities of the numerical model, proposed in Chap-
ter 2, by examining the performance and the physics of the fuel cell under
different scenarios. Some of which include: providing an explanation of the
experimental observations that are presented in Chapter 3, and predict-
ing the performance of the fuel cell under various FEC porous structures,
and various FEC and membrane thicknesses. A comparison between the
FE-DMFC’s and DMFC’s performance is presented.
Chapter 5 summarizes the conclusions and contributions of this dissertation. Possible
future paths in which this study could be extended are also discussed.
Chapter 2
Modeling Approach and Formulation
As current FE-DMFC performance models are all single phase, and few fuel cell multiphase
mixture model (MMM) formulations are single domain, this chapter details a formulation
that solves both of these issues. This chapter is intended to clearly detail:
• the assumptions and their expected impact on the model’s predictions;
• the derivation and implementation of the model; and,
• the challenges and considerations needed to obtain a robust and stable solution.
The proposed approach is expected to simplify the implementation of future fuel cell
models, provide greater insight into the physics of the fuel cell and more realistic predictions
of the fuel cell’s performance. It should be noted that the information in this chapter is
based off of the author’s previous published work [90–92].
2.1 Computational Domain
A schematic of the fuel cell’s computational domain is shown in Figure 2.1, and extends
across the whole MEA and excludes the channels. The reduced area caused by the presence
of the ribs, however, are accounted for at the anode and cathode channel-BL interfaces,
and is treated with an ‘effective channel and rib area’, whereby the anode and cathode flow
27
28
HeCOOHOHCH 66223 OHHeO 22 36623
OHCOOOHCH 2223 223
Flow
ing
Ele
ctro
lyte
Cha
nnel
(FE
C)
Cat
hode
Bac
king
Lay
er(C
BL
)
Cat
hode
Mem
bran
e(C
M)
Cat
hode
Cat
alys
tLay
er(C
CL
)
Ano
deFu
elC
hann
el(A
FC)
Cat
hode
Air
Cha
nnel
(CA
C)
Ano
deB
acki
ngL
ayer
(AB
L)
Ano
deC
atal
ystL
ayer
(AC
L)
Ano
deM
embr
ane
(AM
)
Ano
deFu
elC
hann
el(A
FC)
Cat
hode
Air
Cha
nnel
(CA
C)
y
x
Figure 2.1: Schematic of the computational domain used in the model. The layers enclosedwithin the dashed box are the ones that are considered within the model.
fields are each assumed to be collapsed to one large channel and rib.
This single domain model considers all layers within the domain to have finite thickness
(BLs: 280 μm, CLs: 28 μm, membranes: 183 μm, and FEC: 610 μm), and considers the
domain to be 1D. The 2D form of each equation, however, will be needed to account for the
presence, and introduction and removal of mass at the FEC inlet and outlet. To decrease
computational time, each variable is confined to layers where the transport equations are
physically meaningful, or where information of a particular variable is not known without
solving the governing equations [71]. This will be expanded upon in Section 2.6.
In this work, the pathways shown below are assumed to exist for each species. The fuel
cell that is being modeled is the one that was experimentally tested in this work; its design
29
will be detailed in Section 3.3.
• Methanol is taken to be supplied at the anode, and is transported towards the cathode.
• Oxygen is taken to be supplied at the cathode and is transported towards the anode.
• Water is taken to be supplied at the anode, cathode and FEC inlet and is allowed toflow as governed by the conservation equations.
• Sulfuric acid is taken to be fully contained within the FEC.
2.2 Modeling Assumptions
Many of the assumptions made in this model are similar to those mentioned in Colpan et
al.’s model [35]. However, in the presented model, both the liquid and gaseous states are
considered, also the membranes are no longer assumed to be fully hydrated. The remaining
assumptions are listed below in italics. After each assumption, a discussion on the validity
and impact of these assumptions on the final solution are provided.
1. The fuel cell operates under steady state and isothermal conditions :
Realistically, the fuel cell will operate under transient conditions due to any non-
uniformities within the fuel cell, external disturbances, and obstructions within the
porous media caused by the bubble and droplet motion within the fuel cell. Further-
more, over time the fuel cell performance will degrade due to factors such as: catalyst
poisoning, carbon corrosion, membrane degradation, and MEA delamination [93–95].
A degradation study is outside the scope of this work and will not be examined. But the
model can be considered to look at a statistical steady state, which is often considered
in steady state models and experiments. Furthermore, the fuel cell will also realis-
tically operate under non-isothermal conditions, due to thermal contact resistances,
chemical reactions in the CLs, convective heat transfer from the AFC, CAC and FEC,
30
latent heat from evaporation and condensation processes and joule heating [65,96]. As
such, this assumption will negate any temperature driven transport of species, such as
thermo-capillary action and the “heat pipe effect”, as well as negate any evaporation
and condensation of species [67]. In literature, it has been shown that DMFCs and
PEMFCs tend to exhibit temperature differences, as high as ∼ 5◦C [97–99]. However,it is suspected that the temperature profile within the FE-DMFC will be more uniform
due to the convective heat transfer and large thermal capacitance of the FE.
2. All fluids are ideal and exist in equilibrium with one another :
This is a common assumption in modeling literature due to the complexity of account-
ing for thermodynamic non-ideality, which results in having to determine the spatial
activity coefficient distribution of each species [100]. There is also evidence in DMFC
and PEMFC literature that these fuel cells operate in a state of non-equilibrium, due
to the slow rates of absorption and desorption of water into and out of the electrolyte
phase [101,102]. Furthermore, it is not known if the FE-DMFC is in a state of equilib-
rium. Therefore, for simplicity, the FE-DMFC is treated to be in a state of equilibrium.
This would cause the sorption processes to occur infinitely fast, likely causing an over-
predic